Templates

The Library tab has seven starter geometries. Open one and it becomes a new project — every node, width and port stays editable.

Each preview is drawn from the template’s real channel graph through the same solver the canvas uses, so what you see on the card is the geometry you get.

These are recreations, not reproductions

Every template is built from published device descriptions, with dimensions chosen to print cleanly on a desktop mSLA machine and to sit comfortably inside the default printer profile’s build volume. They are not scaled from any particular paper’s device, and the numbers are not that paper’s numbers.

Where a published geometry cannot be expressed by a planar channel graph at a single Z, the template says so in a caveat and the app shows it in a caution box. Read it before you build on the template.

Choosing

If you want	Start with
Something that works, no thinking	Serpentine mixer
To change one variable at a time	T-junction mixer
To merge two streams gently	Y-junction mixer
Mixing in the shortest path	Herringbone mixer
Long path, no corners	Spiral mixer
Four concentrations from two syringes	Gradient generator
Droplets	Flow-focusing droplet generator

---

Serpentine mixer

16 segments · 3 ports · 215.8 mm of channel · 40 × 60 mm

The workhorse. Two inlets meet at a Y and the combined stream is folded back and forth across the chip six times, giving roughly 200 mm of contact length in a 40 × 60 mm footprint.

Start here if you are not sure what you need. It is easy to print, easy to clear of resin, and the only parameter that really matters is how long you make it.

How it works. At these Reynolds numbers there is no turbulence to help, so mixing is pure molecular diffusion across the interface between the two streams. Length buys time: doubling the path doubles the residence time, and the mixing length scales with the square of channel width.

---

T-junction mixer

6 segments · 3 ports · 51 mm of channel · 36 × 40 mm

The simplest two-stream contactor there is: opposed inlets, a right-angle turn, and 27 mm of straight channel to mix in.

Compact and unambiguous, which makes it the right starting point when you want to change one thing at a time and see what it does.

How it works. The two streams collide and turn together, which stretches the interface at the corner and gives a small mixing head start over a Y. Downstream it is diffusion again, so the straight leg is doing most of the work.

---

Y-junction mixer

5 segments · 3 ports · 47.5 mm of channel · 36 × 44 mm

Both inlet arms curve so they arrive at the junction running parallel. The streams merge as neat co-flowing laminae rather than slamming into each other — what you want for cells, emulsions, or anything else a head-on collision would damage.

How it works. Co-flow produces a clean, flat interface between the two streams and holds it stable down the mixing leg. That makes it the gentlest merge available, and also the slowest to mix, since nothing is stretching the interface.

Note

The gentleness is in how the streams meet, which a one-dimensional model cannot see. The Fluid tab reads the same wall shear here as for a T of the same section and flow, because wall shear follows the channel, not the corner. Do not read equal numbers as equal treatment of your cells.

---

Herringbone mixer

13 segments · 3 ports · 98.3 mm of channel · 30 × 76 mm

Nine chevron legs in three staggered groups. Each group biases the run to one side and the bias reverses between groups — the planar equivalent of the alternating half-cycles in a staggered herringbone mixer.

Mixes in far less length than a plain serpentine, at the cost of sharper corners and a higher pressure drop.

How it works. Repeatedly reorienting the flow stretches and folds the interface between the two streams, so the diffusion distance shrinks geometrically rather than linearly. This is chaotic advection — the same reason folding dough is faster than stirring it.

This is not the published staggered herringbone mixer

The published device works by ridges cut into the channel floor, which are sub-channel features that a planar graph at a single Z cannot express.

This is the planar chevron interpretation: it folds the interface by turning the whole channel rather than by driving a helical secondary flow. Expect useful mixing, but not the original device’s numbers.

If you need the real thing, you need a fabrication route that can produce floor relief — soft lithography, or a printer and resin combination you have characterised well below the channel scale.

See Mixing without turbulence for what the difference means physically.

---

Spiral mixer

24 segments · 3 ports · 249.9 mm of channel · 60 × 60 mm

An Archimedean spiral from a 22 mm outer radius down to 7 mm, with 6 mm between turns. Long path length in a compact square footprint and, unlike a serpentine, no corners at all — nothing to catch a bubble on and nothing to spike the pressure drop.

How it works. Sustained curvature sets up Dean vortices: a pair of counter-rotating secondary flows in the channel cross-section that stir the fluid transversely as it travels. Dean flow strengthens with flow rate, so this is the geometry that gets better as you push it harder.

Note

The Fluid Dynamics tab does not model Dean flow. It sees a long channel of the stated cross-section and reports its pressure drop and residence time correctly — but the extra mixing the curvature buys you is not in any number it shows.

---

Gradient generator

47 segments · 6 ports · 278.2 mm of channel · 60 × 70 mm

Two inlets split, mix and recombine through two levels of branching to produce four outlets carrying a linear concentration series.

The classic tool for dose–response and chemotaxis work: one syringe of buffer, one of compound, four conditions at once.

How it works. Each branch point splits a stream in half and each junction averages two neighbours, so the concentration steps linearly across the outlets. The serpentine between levels exists purely to give each combined stream time to homogenise before it is split again.

The series depends on balanced resistance

The linear series holds only if every branch has the same hydraulic resistance. Widen one channel and the split stops being even — a branch twice as wide carries roughly sixteen times the flow, because resistance scales with the fourth power of the hydraulic diameter.

Check the Fluid Dynamics tab after any edit to this template. The per-segment flow column is where an unbalanced branch shows up.

---

Droplet generator

Flow-focusing droplet generator

6 segments · 4 ports · 62 mm of channel · 40 × 50 mm

A dispersed phase entering from the top is squeezed between two continuous-phase streams and forced through a narrow orifice, which breaks it into droplets.

The throat is built from dimension markers rather than as a separate channel, so you can drag it wider or narrower and watch the neck reflow in 3D. See Dimension markers and lofting.

How it works. Where the streams converge, viscous shear from the continuous phase overcomes the interfacial tension holding the thread together and it pinches off. Droplet size follows the orifice width and the flow-rate ratio, which is why both are the first things to tune.

The throat is near the printer’s limit

0.35 mm is near the limit of a standard 4K printer and will be the first thing the warnings panel flags on a coarse profile. Print a test coupon before committing a chip — see A first test-print protocol.

See Droplet generation for the physics and what the app does not model.

---

What every template guarantees

presets.test.ts asserts that no template opens with a risk-severity warning on the default printer profile. A starter chip that greeted a new user with a red warning would teach them to ignore the warnings panel, and the entire safety model here is advisory.

Cautions are allowed and sometimes deliberate — the droplet generator’s 0.35 mm throat is genuinely tight, and saying so is the point.

Next

• Mixing without turbulence — why these geometries look the way they do.
• Drawing channels — modifying one.